Search results

Question about ultrafineweb dataset preparation for tequila/sherry

Status: Open.

#233 In Tencent/AngelSlim;

· koshostanford opened on Jan 28, 2026

[Eagle-3 Training] Cloud I use this code for Qwen3-235B draft model training. I can<code>t find any config about target model</code>s parallel setting like tp\pp\ep ...

Status: Open.

#230 In Tencent/AngelSlim;

· TheUgliestOfWuli opened on Jan 26, 2026

Benchmarking SpecExit

Status: Open.

#229 In Tencent/AngelSlim;

· dhia680 opened on Jan 23, 2026

hunyuanocr量化问题

Status: Open.

#227 In Tencent/AngelSlim;

· Cliveliew opened on Jan 23, 2026

推理eagles模型时报错

Status: Open.

#219 In Tencent/AngelSlim;

· yancaoweidaode opened on Jan 20, 2026

请问支持eagle v1版本的训练和推理吗

Status: Open.

#189 In Tencent/AngelSlim;

· 2020zyc opened on Jan 7, 2026

draft_vocab_size

Status: Open.

#188 In Tencent/AngelSlim;

· tankpig opened on Jan 7, 2026

datasets

Status: Open.

#183 In Tencent/AngelSlim;

· ggg-s opened on Dec 24, 2025

eagle3 vlm训练,数据预处理过程很慢,而且内存占用很高

Status: Open.

#175 In Tencent/AngelSlim;

· wincle opened on Dec 15, 2025

似乎发现一个bug

Status: Open.

#174 In Tencent/AngelSlim;

· wincle opened on Dec 15, 2025

eagle3训练,模型加载时,没考虑2b小模型没有model.safetensors.index.json的情况

Status: Open.

#173 In Tencent/AngelSlim;

· wincle opened on Dec 15, 2025

How to repreduce the reported results of Qwen3-30B-A3B with EAGLE3?

Status: Open.

#163 In Tencent/AngelSlim;

· Zzsf11 opened on Dec 8, 2025